Spaces:
Runtime error
Runtime error
Delete app.py
Browse files
app.py
DELETED
|
@@ -1,44 +0,0 @@
|
|
| 1 |
-
import gradio as gr
|
| 2 |
-
import torch
|
| 3 |
-
import librosa
|
| 4 |
-
import json
|
| 5 |
-
from transformers import pipeline
|
| 6 |
-
from stitched_model import CombinedModel
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
| 10 |
-
|
| 11 |
-
model = CombinedModel("ak3ra/wav2vec2-sunbird-speech-lug", "Sunbird/sunbird-mul-en-mbart-merged", device="cpu")
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
def transcribe(audio_file_mic=None, audio_file_upload=None):
|
| 15 |
-
if audio_file_mic:
|
| 16 |
-
audio_file = audio_file_mic
|
| 17 |
-
elif audio_file_upload:
|
| 18 |
-
audio_file = audio_file_upload
|
| 19 |
-
else:
|
| 20 |
-
return "Please upload an audio file or record one"
|
| 21 |
-
|
| 22 |
-
# Make sure audio is 16kHz
|
| 23 |
-
speech, sample_rate = librosa.load(audio_file)
|
| 24 |
-
if sample_rate != 16000:
|
| 25 |
-
speech = librosa.resample(speech, orig_sr=sample_rate, target_sr=16000)
|
| 26 |
-
speech = torch.tensor([speech])
|
| 27 |
-
|
| 28 |
-
with torch.no_grad():
|
| 29 |
-
transcription, translation = model({"audio":speech})
|
| 30 |
-
|
| 31 |
-
return transcription, translation[0]
|
| 32 |
-
|
| 33 |
-
description = '''Luganda to English Speech Translation'''
|
| 34 |
-
|
| 35 |
-
iface = gr.Interface(fn=transcribe,
|
| 36 |
-
inputs=[
|
| 37 |
-
gr.Audio(source="microphone", type="filepath", label="Record Audio"),
|
| 38 |
-
gr.Audio(source="upload", type="filepath", label="Upload Audio")],
|
| 39 |
-
outputs=[gr.Textbox(label="Transcription"),
|
| 40 |
-
gr.Textbox(label="Translation")
|
| 41 |
-
],
|
| 42 |
-
description=description
|
| 43 |
-
)
|
| 44 |
-
iface.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|